Norms as a tool for the study of homography

BITTER BITTER WITTER Zz8»A.

BITTER

DO NOT SKIP ANY WORDS. WORK RAPIDLY, BUT LEGIBLY. WORK FROM LEFT TO RIGHT.

Figure 1. Instructions from cover page.
The measures of dominance correlated .953 (df = 190), and the respective stability scores correlated .806. The analysis of regression Unes predicting dominance score from one sex to the other showed six words having significant (>2) standard residuals from both prediction lines (male from female and vice versa). Of the six, four were cases of more extreme dominance in one sex or the other (see Table 1). Only two items were of opposite polarity: “deed” and “iron.”
Table 1
Items That Differ Between the Sexes

Homo­graph	Male		Female
	Dominant Sense	Score		Dominant Sense	Score
Capital	Major	66		Major	90
Charge	Related to Price	54		Related to Price	80
Deed	Legal Document	54		Action	66
Iron	Metal	74		Press Clothes	62
Mint	Flavor	68		Flavor	96
Pit	Hole	72		Hole/Fruit Stone	50

Dominance and Stability
Although we obtained slight sex differences as described above, it did not seem necessary to eliminate any items for our analysis of the general relationship between homograph dominance and stability. Accord­ingly, we combined the performance of the two groups by averaging their respective measures on dominance and stability. The obtained correlation between dominance and stability in the combined sample was r=.813 (df = 190). An examination of the line of best fit pre­dicting stability from dominance revealed an interesting effect. Of the 24 points (12 items by two response types for each) for the heterophones, 22 fell above the regres­sion Une. We therefore analyzed the data separately for the 12 heterophonic and 84 homophonic homographs.
In the case of the 84 homophones, the correlation between dominance and stability (r = .838, df= 166) indicates that knowledge of dominance will enable us to account for 70% of the variance in the stability measure. Similarly, for the 12 heterophones, the obtained correla­tion (r = .845, df = 22) accounts for 70% of the variance in the much smaUer number of cases.
Table 2 presents the norms for the combined male and female samples. Examination of the scores clearly shows that the heterophone items are more stable in meaning than are items of equal dominance.
Regional Differences
Geis and Winograd (1974) have provided norms for Emory University (Atlanta, Georgia) for 50 homographs based on the responses of 68 undergraduates (sex unspecified). Of these 50 items, 45 had been included in norms gathered at Adelphi. However, two words had been eliminated on the basis of pilot work: “sage,” since less than 75% of Adelphi undergraduates could pro­duce a single classifiable association to that string, and “volume,” which had elicited three distinct meanings in a pilot study. Three additional words, “club,” “sole,” and “toll” had produced four, three, and three distinct meanings, respectively, at Adelphi and, therefore, were omitted from the comparison. It should be noted that Geis and Winograd report only the classification of the first two associations in terms of dominance in their table.
A total of 42 words (84 meanings) remained common to the two studies. The obtained correlation (.904) for dominance was quite satisfactory. The regression line predicting Emory scores from the Adelphi sample was computed, and those items that showed a high standard residual (2 or greater) were examined for the nature of the difference. Three such items were discovered: “organ,” “pot,” and “right.” Of these, the word “organ” was not quite as extremely polarized toward “musical instrument” in the Adelphi sample, whereas the other two showed reversal in dominant meaning. While “pot” produced first associations that can be classified as “kitchen utensil” in 77% of the Emory sample, 52% of the Adelphi group produced associations related to “marijuana.” Finally, “right” produced a preponderance (82%) of associations scorable as meaning “correct” in the Atlanta sample, but 65% of the Long Island sample produced responses to “right” as a “direction” (i.e., as opposed to “left”).
The Pittsburgh norms of Perfetti et al. (Note 1), based on a sample of 100 undergraduates, overlapped